Try   HackMD

Deep-Learning-Workshop (12-15hr)


  • Target audience:
    • have some experience in programming
    • knowing how to write for-loop and Class in Python
  • Goal of the workshop:
    • Understand principles of Deep Learning
    • Get familiarized with TensorFlow 2.0 API
    • Multi-GPU and Mixed-Precision Training

Day 1: Overview of AI for Computer Vision

  • The principle of modern AI:
    • Loss
    • Regularization
    • Backpropagation
  • Record Gradients with Gradient Tape
    • code: Linear Regression
  • AI for Computer Vision
    • Brief History of Convolutional Neural Networks (CNN)
    • How Convolutional Layer works
    • code: Convolutional layer's Input/Output

Day 2: Getting Started with TensorFlow 2.0 and CNN

  • Layers: BatchNorm for normalization and Relu for non-linear activation
  • Model construction with tf.keras
    • Sequential API, Model API, Sub-classing API
    • code: VGG, ResNet , DenseNet

Day 3: Multi-GPU & Mixed-Precision Training

  • Horovod for Multi-GPU training
  • Training Speed-up with Automatic Mixed-Precision
  • Introduction to keypoint detection
    • code: facial landmark detection

Image Not Showing Possible Reasons
  • The image file may be corrupted
  • The server hosting the image is unavailable
  • The image path is incorrect
  • The image format is not supported
Learn More →

翁啟閎
Data Scientest @ HonghuTech
chihung@honghutech.com